Build it with an AI agent
Your AI agent can build and run almost all of a Learnt workspace for you: the course, the quizzes, the branding, a separate group for each client organization, invitations and progress reports. You keep the decisions. It does the clicking.
You don't need to be technical. If you'd rather build it yourself, everything below also works in the Course Editor.
What you can ask it to do
- Turn your material into a course. "Turn these notes into modules, one per topic." The agent writes the modules and adds them in order. It can also export the whole course as one markdown file.
- Add quizzes. "Add a five-question quiz after module 3, pass mark 80%." Multiple choice, true/false and select all that apply. Learnt can also draft questions from your modules for the agent to keep or drop.
- Brand it. Name, color, support email, certificate signer and the look of the learner pages.
- Set up a client organization. "Add Ridgeline Realty as a client until June 2027." Each organization gets its own group of learners in the same course, with its own access dates.
- Invite and remind people. It invites learners, resends invitations and sends reminders, and shows you who will get an email before anything is sent.
- Report on progress. "Who at Ridgeline hasn't finished?" Progress per learner, for everyone or one organization, also as a CSV file.
- Handle learner questions. It lists the questions learners asked from module pages and drafts a reply from your own course. It never sends the reply; that stays in your email.
- Run more than one course. Create, rename, archive or switch courses in the same workspace.
What you do yourself
- Start the trial and pay for the workspace.
- Create the API key the agent uses, in the Course Editor under Developer access.
- Connect your own domain, if you use one.
- Upload videos and pictures in the Course Editor. Agents using the Learnt skill can also upload pictures.
How to connect your agent
Pick the route that matches your agent:
- Claude and Claude Code: connect over MCP. See Connect an AI app. Learnt is also listed in the official MCP Registry as
tech.learnt/workspace. - Claude Code and Codex: install the Learnt skill, which teaches the agent the order to build in and adds a helper for pictures and folders of files.
- Any agent that can call a web API: use the Learnt API.
Other apps that accept a custom header, such as Cursor and VS Code, use the same MCP setup, but we haven't tested them. ChatGPT's connectors need OAuth sign-in, which Learnt doesn't offer yet.
Safe by design
- A key reaches only its own workspace. It runs the same checks as the Course Editor, so an agent can't see another client's data.
- Read-only keys. Choose Only read when you create a key, and the agent can report but not change anything. You can revoke any key in the Course Editor, and it stops working on its next request.
- Nothing is emailed without your OK. Invitation and reminder tools refuse to send until the agent confirms you've agreed, so it has to show you the list first. Invitations count against your daily limit.
- Deletes are marked. Tools that delete or overwrite something say so, and most AI apps ask you before running them.
- Errors say how to fix them. Every refusal carries a plain message, a stable code and a link to the fix, so the agent can recover on its own.
Tested with a fresh agent
We test this with a fresh AI agent that has never seen Learnt: it gets only the skill or the MCP server and a key, then does 16 real builder tasks in a throwaway workspace, from building and branding a course to inviting learners and reading progress. We check the workspace afterwards, not the agent's own report. We run it after changes agents rely on, and every month.
It's included in the $99/month workspace. There's no separate API plan.